Sparse seismic data regularization in both shot and trace domains using a residual block autoencoder based on the fast Fourier transform

Abstract

The increasing use of sparse acquisitions in seismic data acquisition offers advantages in cost and time savings. However, it results in irregularly sampled seismic data, adversely impacting the quality of the final images. In this paper, we propose the ResFFT-CAE network, a convolutional neural network with residual blocks based on the Fourier transform. Incorporating residual blocks allows the network to extract both high- and low-frequency features from the seismic data. The high-frequency features capture detailed information, while the low-frequency features integrate the overall data structure, facilitating superior recovery of irregularly sampled seismic data in the trace and shot domains. We evaluated the performance of the ResFFT-CAE network on both synthetic and field data. On synthetic data, we compared the ResFFT-CAE network with the compressive sensing (CS) method utilizing the curvelet transform. For field data, we conducted comparisons with other neural networks, including the convolutional autoencoder (CAE) and U-Net. The results demonstrated that the ResFFT-CAE network consistently outperformed other approaches in all scenarios. It produced images of superior quality, characterized by lower residuals and reduced distortions. Furthermore, when evaluating model generalization, tests using models trained on synthetic data also exhibited promising results. In conclusion, the ResFFT-CAE network shows great promise as a highly efficient tool for the regularizing irregularly sampled seismic data. Its excellent performance suggests potential applications in the preconditioning of seismic data analysis and processing flows.